Correcting time offsets and enclosure-induced measurement distortions in LiDAR–camera systems

Using protective enclosures for multi-sensor perception systems in hazardous environments, such as underground mines, introduces critical measurement errors that are often overlooked in sensor fusion pipelines. This paper addresses two key sources of these errors: temporal misalignment between sensor data streams and refractive distortion at the optical interface. To address temporal misalignment, an observation-based temporal calibration method is introduced, where a rotating flywheel observed by both sensors provides a direct estimate of the inter-sensor time delay, reducing the timing error from approximately 32 ms to below 3 ms without external triggering or synchronisation hardware. To account for enclosure-induced distortion, LiDAR beam refraction through a 6 mm polycarbonate dome is modelled using analytical ray tracing, validated by physics-based simulation, compared with adapted refractive baselines, and corrected in real time using a lookup-table approach with negligible computational overhead. Enclosure-aware camera intrinsic and LiDAR–camera extrinsic calibrations are then performed with the enclosure in place, treating the sensors and enclosure as a single optical system. Experimental results show that LiDAR refraction correction improves dimensional accuracy by 43% at 5 m, enclosure-aware intrinsic calibration reduces the mean reprojection error by approximately 58%, and extrinsic calibration reduces projection error from approximately 10 pixels to 2 pixels. Ablation and repeatability analyses further confirm that accurate fusion requires the combined use of LiDAR refraction correction, enclosure-aware camera calibration, and enclosure-aware extrinsic calibration.